View
7
Download
0
Category
Preview:
Citation preview
Original Article
Development of a new method for
ATFCM based on trajectory based
operations
Dany Gatsinzi, Francisco J Saez Nieto, Irfan Madani
Abstract
This paper discusses a possibility to evolve the current Air Traffic Flow and Capacity
Management towards a more proactive approach. This new method focusses on reducing
the expected probability of ATC intervention based on "hot spot" identification and
mitigation at strategic level by applying subliminal changes on the times of arrival at the
crossing or merging points (junctions). The concept is fully aligned with the Trajectory
Based Operation principles. The approach assumes that the changes on the times of
arrival only demand small speed changes from the involved aircraft. In this study, the hot
spots are defined as clusters of aircraft expected to arrive to the junctions. Two aircraft
are said to be in the same cluster if their proximity and closure rate are below a given
threshold.
School of Aerospace, Transport and
Manufacturing, Cranfield University, UK
Corresponding author:
Dany Gatsinzi, School of Aerospace Transport
and Manufacturing, Cranfield University,
Building 83, College Rd, Cranfield MK43 0AL,
United Kingdom.
Email:Dany.Gatsinzi@cranfield.ac.uk
Tel:+447941514458
Some exercises are proposed and solved by applying this method. The obtained results
show its ability to remove the potential conflicts by applying simple linear programming.
This approach seeks to change the current capacity limiting factor, established by the
number of aircraft occupying simultaneously each sector, to another parameter where the
level of traffic complexity, flowing towards junctions, is identified and mitigated at
strategic level.
The speed changes, used as the control variable and computed before or during the flight,
are designed to provide an adjustment on aircraft’s Required Time of Arrival at the
junctions in order to have a de-randomized and well-behaved (conflict free) traffic. This
will enable improvements in airspace capacity/ safety binomial.
It is recognized that this measure alone is unable to produce a conflict free airspace, and
then other collaborative and coordinated actions, such as adjusting and swapping
departing times at the departing airports (before the Aircraft are taking off), offsetting
some flights from nominal route, and allowing multi-agent separation management
(while they are in flight) should be applied together with this method.
Keywords
ATFCM, TBO, CD&R, DCB, junction, TOA, hot spot, complexity
1 Background The Air Traffic Flow and Capacity Management (ATFCM) has been the object of
intensive research, due to the sustained growth of air traffic. It is a required service for
managing the balance between air traffic demand and airspace capacity, defined by the
ATC operational resources. Since its inception, the Air Traffic Management (ATM)
system has undergone several significant changes from a primitive form, based on simple
operational rules, to a complex network supported by sensors, communication and
control subsystems.1 In addition, the evolution of the technical enablers such as radio
communication, navigational aids (Navaids) and radar systems which revolutionised the
way ATM system was operating in 1950s, still forms the backbone of the current
system.1 Derived from those supporting technologies, today’s ATM system is still
basically based on a network of fixed routes and ATC sectors.
To achieve the required levels of safety for air traffic across this network, a
human supervisory control system is applied. The process is based on air traffic
controllers (ATCOs) detecting and solving aircraft (A/C) conflicts within each ATC
sector in a reactive manner. The conventional airspace organization, which is constituted
of a network of fixed routes, sectors and nodes, enables the ATCOs to have a priori
information on the potential A/C conflicts since they are typically located at the
crossing/merging points of the route network. The number of simultaneous A/C that can
be handled within a sector is limited; saturating the ATCO’s capacity. To maintain an
acceptable controller workload, a maximum allowable number of simultaneous A/C
within a sector is defined, regardless of their particular complexity. The result is that the
current airspace capacity then strongly depends on the individual capacity of each ATC
sector. The workload being experienced by ATCOs is the main limiting factor of the
whole airspace capacity; and the current ATM operations are commonly referred to as
“airspace based operations”.
At present, airspace resources are used to attend demand, while they suffer from
idleness and saturations that foster regulations and holdings, which causes inefficient
flight trajectories.
The above-mentioned situation is more significant in high traffic density airports
and ATC sectors in Europe and it results into unnecessary delays. According to the
Eurocontrol Performance Review Report (2016),2 considering only the ATFM en-route
delays, an average of 51seconds delay per flight and a total of 8.7million minutes delay
were recorded in the ECAC area, resulting in en-route ATFM delay costs estimated at
8.7million Euro. In addition to these substantial delay costs, the projected rising demand
for air travel has the potential to further increase air traffic congestion and reduce the
ATM operational safety and efficiency.3 With the increase of traffic congestion in ATC
sectors, the Air Traffic Controllers’ workload, which mainly increases due to increased
number of tactical actions required to remove conflicts between aircraft also increases,
limiting the number of operations that can be attended safely and efficiently by the
controller.4
To overcome these limitations, SESAR, a European ATM joint-research
program,5 as well as its counterpart NextGen in the United States,
6 are actively
researching and proposing procedures to minimise the problems. A key common pillar of
both projects is the “Trajectory Based Operations (TBO)” based on the 4D trajectory
concept, where A/C are expected to fly closely to their own preferred trajectories whilst
meeting the performance requirements defined associated to the procedure flown. Under
this TBO principle, all ATM stakeholders will collaboratively define these trajectories.
Constraints associated to ATFCM will then be solved under the principle of obtaining
trajectories that suits users’ internal business objectives.7
The aforementioned performance requirements, which are defined in ICAO’s
Performance Based Navigation (PBN), allow A/C to navigate their flights following a
planned centered-line trajectory accurately either under “Free Routing” (FR) airspace or
following the conventional route structure. The application of FR changes the current
fixed and well defined “crossing points”, where most of the conflicts are expected to
occur. In FR, the number of junctions is increased and then, for the same population, the
number of aircraft crossing them will be, on average, reduced. However, the complexity
of the airspace might be increased. In addition, the workload for the air traffic controllers
to detect the conflicts, as they are broadly distributed, could be increased.
The air traffic flow management optimization problems have been intensively
addressed over the years. The models developed to adapt demand and capacity can be
classified as:
Macroscopic approaches ;
Microscopic approaches.
Currently, macroscopic approaches are usually applied to wide geographical areas
providing corrective measures before the A/C is flying. They use aggregated flight plans
as the input, and the ATC sectors’ and airports’ capacities as the “limiting factor”.
Throughout a demand capacity balancing (DCB) dynamic exercise, the initial flight plans
can be directly approved or “regulated”. The regulation action is then the available
“control action” and in Europe, this usually implies limiting the maximum rate of aircraft
entering either a regulated volume of airspace or airport by modifying flight plans, agreed
between Flight Operating Centres (FOC) and Network Manager (NM), which may either
involve rerouting (new A/C trajectory) or assigning new take-off times through ATFM
time slots.
The ground holding is the most common regulation criteria. The “ground delay
problem (GDP)” seeks to minimize the sum of airborne and ground delay costs when the
demand exceeds the allowed capacity. The GDP problem has been studied in a simple
airport capacity constrained scenario, 8
and in a more complex scenario involving
capacities of a network of interconnected airports and connecting flights while
considering cascading effect of delays on flights in the network. 9 In both scenarios, the
airspace (ATC Sector(s)) capacity was assumed unconstrained, and speed control and
rerouting were not considered as possible solutions. In a congested airspace such as that
over Europe, these assumptions are idealistic.
To remove this unrealistic airspace capacity assumption, a deterministic
programing model was developed to solve the ATFCM problem considering the full
network capacity constraints (airports’ and en-route sectors’ capacity constraints). 10
For
each A/C in the network, the model finds optimal departure and sector occupancy time
that minimises the total network delay. The authors concluded that the problem is NP-
hard and suggested a heuristic which proved to be computationally efficient. This model
however, also does not allow for rerouting and speed changes as control variables. A
more complete model that allows rerouting has been recently developed. 11
This model
has been validated on large-scale traffic instances of the size of US National Airspace
with viable computational times. Nowadays, this model is the most complete
macroscopic description of the ATFCM problem.
Currently, these macroscopic approaches for ATFCM problems are sufficiently
powerful to address large-scale scenarios in which A/C trajectories are represented as
aggregated flow, but they are not able to account for the effects of potential conflicts that
may arise from individual planned trajectories on the ATCO’s workload.
Flow management microscopic approaches are currently focused on conflict
detection and resolution (CD&R), including sequencing, merging and metering when
competing for limited resources (such as runways and merging points). The goal is to
provide conflict free trajectories for all involved A/C. CD&R algorithms are seldom
applied as ATFCM operational solutions to address the airspace capacity problem,
whereas they are inherently the common ATCO’s reactive/tactical solutions. A
comprehensive survey of CD&R algorithms is provided in Kuchar and Yang. 12
Apart
from ground or air holdings, three well-known techniques exist that can be applied to
remove A/C conflicts; heading, flight level and speed changes. Currently however,
conflict resolution by speed control is rarely applied. This is mainly due to the limited
possible speed changes, compatible with A/C performance. Speed control requires
significant anticipation to be efficient, which is not the case for heading or for flight level
changes. Heading and flight level changes can quickly avoid loss of separation than speed
control. Consequently, these techniques are more intuitively suitable than speed control
and they are paramount under time pressure (reactive situations) to resolve conflicts.
This paper focuses on reducing the probability of ATC intervention to resolve conflicts
far in advance (at strategic ATFM level) when there is enough anticipation time suitable
for speed control actions. These speed changes will be applied to the flight plans (RBT
level) as time constrains, for instance, issued in form of RTAs.
A hybrid CD&R model combining speed control and flight level changes has also
been proposed by Vela et al., 13
formulated as a Mixed-Integer Linear programme
(MILP), some encouraging results were obtained in the paper. However, the model was
implemented with a consideration of the airspace size of only one ATC sector. One of the
reasons why speed control was poorly investigated was the lack of accuracy of trajectory
predictions. However, the new developments of CNS avionics technology such as the
FMS CTA function opens the door to the application of speed control based conflict
resolution models.
More recently, several European ATM projects have focused on developing tools
to provide strategic trajectories de-confliction functions, exploring the impact of applying
time constraints at given waypoints. These tools rely on a key pillar of SESAR Target
concept, the “Trajectory Based Operations (TBO)”. TBO is based upon the 4D trajectory
concept, which describes for each flight, its intended trajectory, defined with a required
accuracy associated to all three spatial coordinates (3D) and target times to specific
waypoints.14
The 4DT management evolves out of a collaborative layered planning that
spans through different phases. The initial RBT, which is the trajectory airspace user
agree to fly and the airports and ANSPs agree to facilitate, is an outcome of the demand
capacity balancing process.14
This RBT is a reference, but it is not required to adhere
with specific margins in time or space, unless some initial constrains are added to its
specification. The RBT update and revision processes are an integral part of the SESAR
concept. Initial RBT is closely related to the ICAO EFPL concept, representing an
extension to the conventional ICAO flight plan. It provides the 4D-trajectory description
as computed by the flight planning system to generate the operational flight plan; it also
provides performance data describing the climbing and descending capabilities specific to
the flight.
The improved accuracy associated to the 4DT increases trajectory predictability,
this has led to innovative studies on new trajectory de-confliction methods to allow better
traffic management processes at different planning layers.
One of them is the “En-Route Air Traffic Soft Management Ultimate System
(ERASMUS)”.15,16
Its main objective was to develop a tool to improve A/C ATC
strategic de-confliction by generating a conflict-free trajectory segment of 15 minutes
look-ahead time for each flight. This was achieved through the Trajectory Control by
Speed Adjustment (TC-SA) function, which proposes real-time in-flight adjustments of
the RBT based on minor (-6%, -6%) speed adjustments in order to reduce the current
ATC workload. These speed changes are referred to as “subliminal”.
The ERASMUS strategic de-confliction process assumes the availability of
accurate 4D trajectory predictions, where each aircraft downlinks to ground a 4D
trajectory every 2 minutes for a look–ahead time of 20 minutes. According to the existing
4D trajectory constraints, this ground processing system is able to update its picture of
the traffic situation every 3 minutes and then, its conflict detection and resolution
algorithms detects whether or not there are conflicts for the next 15 minutes trajectory
segment of each aircraft. For the detected conflicts, new trajectory clearances for the next
15 minutes window are up-linked to the aircraft in terms of CTO/CTA. These time-
constraints are achieved by minor speed adjustments to the aircraft speed profile. The
intention is to adjust the speed for each pair of A/C in conflict during the 15 minutes time
window in order to increase their separation to 7NM, which implies an additional 2NM
uncertainty buffer at the conflict point.15
The efficiency of the ERASMUS tool was validated through fast-time simulations
(FTS), where the TC-SA resulted into 80% reduction in the number of A/C potential
conflicts handled by controllers, and allowing ATCOs to handle a 70% traffic increase
without any increase in their current workload.17
Based on these results, SESAR decided to investigate this concept further,
through the so-called Trajectory Adjustment through Constraint of Time (TRACT)
service.18
It is expected that TRACT will be developed to manage early (25 minutes look-
ahead time) conflicts identification. Just like TC-SA function, TRACT assess expected
conflicts within its look-ahead time and tries to resolve them by automatically issuing to
the appropriate A/C, CTO constraints over the conflict point, achieved by minor speed
adjustments in order to decrease the ATCO workload due to conflict assessment and
monitoring.
It was concluded that a TRACT solution is only possible to conflicts involving
i4D-capable aircraft and thus, it is well known that several enablers for TRACT are not
available yet. 19
For this reason, TRACT is now being studied along with other automated
support tools for conflict detection and resolution. 19
All these tools will provide services
with different look-ahead times. However, they all lead to the same ultimate objective of
eventually reducing the complexity of traffic in order to reduce controller’s workload and
increase capacity.
The above-discussed tools are automated tools providing strategic detection and
resolution of 4D trajectory conflicts at execution/in-flight phase. They are said to provide
strategic detection of conflicts because they can detect conflicts at a longer look-ahead
time horizon than a typical detection look-ahead used by the controllers.
In this paper, the method proposed is applicable at pre-flight DCB timeframe, and
is considered as a dynamic DCB /Short-Term ATFCM Measures (STAM) as discussed
later in the paper.
2 ATM nodes, links and junctions’ topology
ATM operational network can be seen topologically as a set of fixed nodes
(internal or external), directed links among nodes and links intersections (see Figure 1).
Each internal node (i, j...) is, simultaneously, the sink and source points of traffic flow
and they are representing the physical volume of airspace occupied by a Terminal
Manoeuvring Area (TMA). The external nodes (k, l...) are also considered as sinks and
sources of the traffic, but they are representing entry/exit points of the airspace under
consideration.
Figure 1.Airspace topology
An internal node (i), shown in Figure 1 as a small square, has outbound traffics
represented by exit links (from i to j) departing from any centre of each square’s side, and
inbound traffics (from j to i) represented entering by any square’s corner. The external
nodes are shown as circles, which also have outbound and inbound traffics, as shown in
Figure 1.
Links represent planned A/C trajectory tracks, where the (unidimensional)
continuity principle will be applied along them if they do not arrive to any junction.
Finally, junctions are dynamic or fixed locations where two or more links are expected to
converge. An intersection of links will only be considered as a junction if it is “active”,
that is; when a set of two or more A/C are expected to arrive within a small and well-
defined time interval limit among any pair of them. These junctions might or might not
be co-located with waypoints (WPs). In this research, WPs are classified in two different
types: first, as airspace 3D points where some changes on the velocity vector are
expected or, second, points set when a new operational condition applies.
Figure 1. Junction geometries
In general, the geometry of a junction20
can be defined by their physical
intersection. The junctions could have m incoming links and n outgoing links (see Figure
2). When n=m, it is considered a crossing point whilst when m>n, it is named as a
merging point. The nodes of the Air Traffic Service (ATS) route network are usually
merging points. Finally, when n<m, the point is referred to as a distribution or a fork
junction.
In this paper, it is assumed that outbound traffic (all flows emerging from node i
towards all other nodes j, qij) and inbound traffic (all flows arriving from all nodes j
coming to node i, qji) satisfies the limiting throughput criteria (QIi) and (QOi). Thus,
considering all nodes (N), the following equations can be stated for each node (i):
𝑄𝐼𝑖 ≥ ∑ 𝑞𝑗𝑖
𝑁−1
𝑗=1,𝑗≠𝑖
QOi ≥ ∑ qij
N−1
𝑗=1,𝑗≠𝑖
(1)
Where, QIi and QOi are a priori known, possibly time dependant, maximum allowed flow
values for each node (i).
Managing the TMA capacities in terms of inbound and outbound maximum flows
(QIi and QOi) supported by E-AMAN/DMAN is considered as boundary conditions of
the problem. For ATFCM purposes, all the required information from these nodes is
provided for the above criteria. In other words, all the following discussion refers to
airspace beyond the limits of the TMAs borders.
The first equation in Equation (1) also applies to the (active) junctions. That is to
say; the whole maximum arriving traffic to the junction (m) shall be equal or smaller than
the junction inbound flow capacity (𝑄𝐼𝑚). This capacity limit is particularly important at
merging points (as when they are the entry points at the nodes), where the incoming
traffic is confined into higher density outbound routes. The required “time to arrival
change” raises very fast (towards infinite) when inbound traffic reaches the inbound flow
capacity (𝑄𝐼𝑚).
As it is discussed later in this paper, an inbound flow capacity is very sensitive to
TOA uncertainties. This condition implies that when confining air traffic in a fixed
conventional ATS route network and maintaining the current levels of A/C TOA
uncertainties, the proposed new ATFCM mechanism becomes unrealistic. Then, this
paper postulates that most of the A/C are flying following free routing (FR) airspace.
Free-routing airspace (FRA) enables direct trajectories and the RBT will improve their
predictability.14,21
In fact, flight plans will be more distributed over the entire airspace
and then the traffic density will be generally more homogenously distributed (except
those in the nodes entry points), then the expected probability of ATC interventions
would be smaller. Therefore, the number of junctions will be higher and the average
number of conflicts per junction will be reduced. This also implies that the ATCO’s
activities will shift from a situation of attending to a high number of conflicts,
concentrated in specific junctions, to managing less conflicts per junction distributed over
a much larger portion of the airspace.
As discussed later in this paper, when the inbound conflicted traffic exceeds the
junction’s inbound flow capacity (𝑄𝐼𝑚) established to be 6A/C per hour , the required
“TOA/Speed changes” to remove conflicts at junction becomes significantly high and
may not be realisable by the A/C. To achieve realistic speed changes compatible with
aircraft performance,22
it is then postulated that most of the A/C are flying following free
routing airspace (FRA), then the average traffic density per junction (except those in the
nodes entry points) is low. For crossing point junctions within this context, the normal
situation will then have an actual traffic flow below its limit. Under these circumstances,
the situation can be dealt with as a traffic de-randomization problem (by speeding up and
slowing down arriving traffic) to achieve a minimum “safe” time interval among all A/C
arriving at the junction.
3 Conflict and hot spot strategic characterisation
Hot spots or A/C clustering consists in identifying groups of closely spaced A/C
expected to arrive at a junction. The clustering process isolates a set of A/C that will be
involved in multiple conflicts, close in time of arrival at a junction, ensuring that solving
conflicts within each cluster separately, will not generate any new inter-cluster
conflicts.23
Cluster identification can be performed by using traffic complexity metrics, 20
or by aggregating space and temporal induced effects to all pairwise conflicts. In this
paper, the latter criterion is used, determining the conflicts based on their expected TOA
separation at the junctions, complemented with a spatial indicator regarding the traffic
complexity for the conflict’s surrounding area.
The method proposed in this paper is dealing with short-term ATFCM measure
(STAM) issued short time before departure, during the pre-flight DCB timeframe, by
including time constraints specifications into the initial RBT, or extended flight plan
(EFPL), to strategically and pro-actively remove expected conflicts at the merging points
(junctions). This is achieved by incorporating minor adjustments in A/C Time of Arrival
(TOA) at conflicted junctions. The proposed method is then acting at pre-departure
phase, where EFPL provides more accurate 4D flight profiles description. Then the minor
A/C TOA adjustments at hotspots to resolve expected conflicts should enhance the
overall process of traffic management.
A key pillar required for the implementation of the proposed method, is to
consider the A/C trajectories as the basis for both, the planning and execution phases of
the ATM. These A/C trajectories must be continuously managed during their whole life
cycle, since shared as SBT until fully completed in block time.
To determine the required TOA adjustments (achieved by minor speed changes)
for removal of expected conflicts at the junctions, this paper discusses about the effect of
the different sources of uncertainty in A/C TOA at junctions assuming pre-flight planning
layer phase, the STAM time window, where a given probability of ATC intervention in
resolving conflicts at junctions is established. The proposed method is then considered as
a dynamic DCB /Short-Term ATFCM Measure (STAM).
It can be considered as complementary to another STAM currently studied under
the SESAR 2020 exploratory research project, named “Cooperative departures for a
competitive ATM network service (PARTAKE)”. 26
PARTAKE project is focussing on
improving the air traffic dynamic demand capacity balancing, using as a window of
opportunity, the prompt identification of “hot spots” at network level and re-adjusting the
estimated take-off times within the assigned nominal CTOT margins, and rearranging the
departing sequence of aircraft at the involved airports to remove those hotspots. The
proposed method in this paper and PARTAKE project are then somehow complementary,
that is, if the computed TOA adjustments at junctions require speed changes that are
unrealistic, PARTAKE can help-by adjusting the A/C’s estimated take-off times.
Pairwise foreseen conflicts can be identified either by computing the expected
separation at their closest point of approach (CPA) or by determining their expected time
separation at the junction. Any A/C following predefined flight plans, typically defined at
strategic level, usually have relevant along track unpredicted deviations, leading to strong
variations on CPA location and distances. On the contrary, any A/C that usually has a
well-defined 2D (or even 3D position, leading to a well-established 2D or 3D junction)
will have uncertainties on their TOA. Hence, a definition of planned “time interval”
between arriving A/C at a junction has been chosen to identify and to mitigate foreseen
conflicts.
When minor speed changes are introduced to the initial flight speed profiles in
order to remove conflicts at junction, the same proximate percentage of A/C’s TOA at
that junction will be modified. For instance, for an A/C flying at 400 knots for a flight
distance of 134NM, if a speed change between 5% to 10% is applied, the A/C’s TOA at
junction will be also modified by 5% to 10%, equivalent to one to two minutes for every
20 minutes of A/C flight time.
In this paper, every foreseen conflict at any junction among all A/C is identified at
strategic level and the required optimised TOA changes at this junction are computed. A
Simple random sequence is used to generate initial time separation interval (τ0) between
any two consecutive aircraft at the junction before TOA/ speed changes are applied. This
initial time separation interval (τ0) defined in Equation (24) is generated within the range
of [0-9] minutes following a uniform distribution. This range of [0-9] minutes implies
that each aircraft will initially require ATC intervention since we established 9 minutes as
the minimum required separation at junction to reduce ATC intervention as discussed
later in the paper. In other words, τ0 represents the expected time interval between
consecutive arrivals at the junction according to their flight plans. The new speeds
applied to achieve TOA changes may induce conflicts at other junctions of involved A/C
trajectories. Therefore, removal of a particular conflict demands an analysis of the whole
set of junctions along the whole set of involved trajectories to take into account this
interaction.
Additionally, at each crossing/merging point where the expected A/C TOAs have
been changed, an analysis of complexity in the surrounding traffic should be performed
by considering the A/C density and their intrinsic complexity, derived from a metric
based on the ratios between their estimated velocities and distances to the junction.
Based on the two (temporal and spatial) criteria described above, for each (active)
junction it is determined in this paper: First, whether the involved A/C have temporal
(along their trajectories) impact (inducing other conflicts) over other intersected
trajectories, and second; the complexity of the surrounding traffic is determined.
Figure 3 describes the twofold visions of conflict resolution impact: The vision
on impact from along trajectories induced conflict and the vision on the complexity of the
surrounding area.
Figure 2. Spatial and temporal impacts for A/C conflict resolution.
4 Approach
4.1 Encounter removal for subliminal TOA changes based on speed control
At flow and capacity management level, the speed changes should minimise the
expected probability of ATC intervention, providing estimated conflict free flight plans,
but some additional criteria should be added to provide information about the
characteristics of the perturbed area (referred in the Figure 3 as conflict surrounding
area).
The probability of 10-5
is set as the targeted probability for ATC intervention in
resolving potential conflicts at any junction, obtained considering all uncertainties in the
flight plan, as described in the RBT. This probability is derived from the total uncertainty
in the A/C time of arrival (TOA) at junction and the resulting required minimum safe
time interval between any two consecutive aircraft. It is discussed later in this paper that
for the derived total expected standard deviation ( 𝜎𝑇 = 1.5 𝑚𝑖𝑛) in the TOA at junction,
and to achieve an expected probability of ATC tactical intervention of 10-5
, it requires a
minimum safe time interval (𝜏𝑝 = 9 minutes) between the expected TOA for any two
consecutive A/C. Therefore, 9minutes interval is used in this paper as the required
minimum time interval at the junction, when issuing time constrains in the RBTs.
For the proposed approach, incompatibilities of trajectories interdependencies are
determined at strategic level, when issuing the RBT, as proposed in TBO, based on a
given probability of ATC interventions.
The correlation between the geometry of links at junctions and ATCO workload
has been previously acknowledged.24,25
Indeed, understanding this geometry for foreseen
conflicts is crucial in determining the most suitable technique (departing time,
heading/2D-offset, and flight level and speed changes) to resolve it. Therefore, this paper
recognises that subliminal time changes by speed control could not always be the
distinctive technique to solve certain potential conflict geometries, demanding the
concurrency of the others to cope with, at different stages of the flight plan’s lifecycle.
For each (active) junction, the following parameters are relevant to characterise a
conflict:
1. The time interval among the expected TOA of all A/C and their associated
uncertainties;
2. The A/C navigation performance within the links;
3. The angle between the A/C trajectories centerlines at the junction;
4. The expected groundspeed for each A/C;
5. The expected relative altitude of each A/C before and after the crossing point.
Minor speed control technique to remove potential conflicts, shall consider these
different encounter geometries.
4.2 Modelling approach
Recently, new classes of modelling approach in ATFCM based on either
Lagrangian or Eulerian dynamics have arisen.26,27
Lagrangian ATFCM models consider
the dynamics of individual A/C; they can therefore, be seen as trajectory-based models.
On the contrary, Eulerian models usually are macroscopic in nature. They are aggregated
models, which consider the flow rates and densities in control volumes but do not track
individual A/C trajectories.27
For Lagrangian models in ATFCM, the dimension of the problem grows
proportionally with the number of A/C considered, and each individual trajectory must be
precisely propagated in the network. Hence, Lagrangian models in ATFCM involving
large numbers of A/C are generally computationally intensive.26,27
Eulerian models on
the other hand, focus on the aggregate properties of the air traffic network and hence,
have lower fixed dimensions. As a trade-off however, the Eulerian models are generally
not able to provide detailed information about each individual A/C.27
This paper provides a combined approach that, to some extent, uses both the
Lagrangian and Eulerian views in a complementary manner. This approach allows an
Eulerian aggregate optimal traffic flow to analyse local areas and Lagrangian knowledge
of individual flight trajectories. The geometric characterisation of junctions is used for
tracking the trajectory of each A/C to identify any conflicts and for providing required
speed changes of each A/C’s trajectory. Each potential conflict among A/C is
characterised by the time separation at their active junction. An aggregated local Eulerian
vision is used as a metric to measure the complexity of the “disturbed” area where the
conflict is located.
A time separation at the junction is used to formulate a cost function for conflict
resolution, which relies on a linear programing solver that seeks to minimize the total
A/C speed changes required for conflict resolution while maintaining their total flight
time as a constraint.
In order to implement our conflict resolution model on realistic air traffic
scenarios, we treat the required A/C separation by speed control as an optimization
problem that has the following objectives:
1. Maximizing the total number of expected conflicts removed through subliminal
speed changes;
2. Minimise the total amount of speed changes used for conflict resolution.
The minimum required time separation at a junction point is determined by
considering the cross-track, along-track and scheduling uncertainties using a statistical
method.
4.3 Linear programming optimisation problem for a pairwise crossing junction (n=m=2)
This section discusses a simple case where only two A/C and one crossing point
are considered.
Figure 4 picturises the distance vs.time evolution to and from a crossing point of
two A/C from their initial positions at points A and C respectively. Their initial
trajectories are represented by solid lines, connecting to their next waypoints (B and D,
respectively) after the crossing point. In this example, A/C from point C is being the
second arriving A/C at the crossing point. The proposed speed changes involves slowing
down the second arriving A/C (A/C at point C initially) up to the crossing point, and then
flying above its nominal speed from this point. On the contrary, the first A/C arriving at
the crossing point (A/C at point A initially) is doing the opposite. This approach provides
a good platform to formulate the initial linear optimisation cost function to determine
optimal speed changes for each A/C before and after the junction, subject to time
constraints.
Two digit subscripts are used in Figure 4. The first digit subscript refers to
situations before (denoted by 1) and after (denoted by 2) the junction. The second digit
subscript refers to the involved A/C. Finally the tilde symbol (‘) refers to modified
variables (speed or time) over the nominal ones (without tilde).
Figure 4.Distance /Time graph for A/C evolution to and from the junction
For the nominal and adjusted speeds (constant or average speeds are assumed)
𝑉11𝑎𝑛𝑑 𝑉11′ of A/C1, the speeds are defined respectively as:
𝑉11 =𝑑11
𝑡11; 𝑉11
′ =𝑑11
𝑡11′
The speed change can then be expressed as:
∆𝑉11 = 𝑑11(1
𝑡11′ −
1
𝑡11) ≈
𝜏1𝑑11
𝑡112
∆𝑉11 = 𝑉11
𝜏1
𝑡11
Where 𝜏1 is the time in advance of the speeded up A/C arriving at the junction
and let us call 𝑟11 its rate of time or speed change defined as:
𝑟11 =𝜏1
𝑡11
That can be written as well as relative speed change:
𝑟11 = ∆𝑉11
𝑉11=
𝜏1
𝑡11 (2)
Similarly from Figure 5, it can also be derived for the second A/C:
𝑉12 =𝑑12
𝑡12; 𝑉12
′ =𝑑12
𝑡12′
∆𝑉12 = 𝑑12 (1
𝑡12′ −
1
𝑡12) ≈ −
𝜏2𝑑12
𝑡122
∆𝑉12 = −𝑉12
𝜏2
𝑡12,
The ratio of time/speed change (𝑟12) is then:
𝑟12 = −∆𝑉12
𝑉12=
𝜏2
𝑡12 (3)
Following the similar processes, the required speed changes beyond the junction
are obtained as follows:
For A/C 1:
∆𝑉21 = −𝑉21
𝜏1
𝑡21,
𝑟21 = −∆𝑉21
𝑉21=
𝜏1
𝑡21 (4)
And for A/C 2:
∆𝑉22 = 𝑉22
𝜏2
𝑡22,
𝑟22 =∆𝑉22
𝑉22=
𝜏2
𝑡22 (5)
The postulated objective function (to be minimised) is then the aggregation of
“ratios of change”, weighed by their distances, that is:
J= [ 𝑟11𝑑11 + 𝑟12𝑑12 + 𝑟21𝑑21 + 𝑟22𝑑22] (6)
And they are subjected to the following constraints:
𝜏1 + 𝜏2 = 𝜏 , achieving final aditional required time separation (𝜏) (7)
𝑟11, 𝑟12, 𝑟21, 𝑟22 ≤ 0.0𝑋, speed changes below 𝑋% (8)
𝑑11
(1+𝑟11)𝑉1+
𝑑21
(1−𝑟21)𝑉1= 𝑡11 + 𝑡21 → 𝑡21𝑟11 − 𝑡11𝑟21 = 0, unchanged total time 𝐴𝐶1 (9)
𝑡22𝑟12 − 𝑡12𝑟22 = 0, unchanged total time AC2 (10)
All the parameters required in the above optimization model are known except the
minimum required time separation between the two A/C at the junction (𝜏). This
parameter depends on the uncertainties of A/C’s TOA and it is described later in this
paper.
4.4 Example 1: elemental conflict removal by applying TOA subliminal changes
This example considers a basic scenario where the above linear optimization
model is used to compute the required optimised speed changes for each A/C before and
after the joint for a pairwise junction. An ideal condition of no wind is assumed. Only
two flights flying following the great circle are considered: BA560 from Oslo to Lisbon
(ENGM-LPPT) and TAP761 from Rome to London (LIRA-EGLL). The main
characteristics of the analyzed initial flight plans are presented in Table 1:
Table 1. Main characteristics of the analyzed initial flights plans
Flight A/C Flight
Distance
(NM)
Cruise
Speed(Knots)
Flight time
(Hrs)
ENGM-
LPPT(BA560)
A320 1537 453 3.4
LIRA-EGLL
(TAP761)
A320 820 440
1.86
The simulation exercise is performed using ATC2K simulator software.28
The
software uses a simulated Control Working Position (CWP) as the Human Machine
Interface (HMI) for controlling the air traffic. It also presents a Conflict and Risk Display
(CARD) feature which displays a table containing all potential conflicts between A/C
pairs. The CARD includes: The predicted CPA time, the identity of involved A/C and
additional highlights of the route segments in conflict. The situation before applying
speed changes is shown in the left picture of Figure 5 whilst the same situation after
applying TOA changes is shown in the right picture.
Figure 5.Conflict removal applying optimized TOA changes for 2A/C
In this example, applying the TOA changes is considered sufficient to resolve the
conflict where speed changes ensure that the A/C TOA minimum interval is one minute.
The outcomes of the linear optimization are summarized in Table 2:
Despite these promising results, the used strategic identification and removal of
potential conflicts relies only on determinist assumptions and, therefore, conflict-free
trajectories with respect to non-deterministic factors cannot be ensured. This is simply
because the approach used in the example has omitted the characterisation of
uncertainties and might have generated unrealistic solutions.
Table 2. The outcomes of the linear optimisation in Example 1
Flight Speed change before junction
Speed change after junction
ENGM-LPPT(BA560) 4𝑘𝑡𝑠 (0.81%) 1𝑘𝑡𝑠 (0.25%)
LIRA-EGLL (TAP761) 4𝑘𝑡𝑠 (0.92%) 0𝑘𝑡𝑠 (0.04%)
4.5 Deriving the minimum required time separation at the junction (𝜏)
Pairwise conflicts are usually identified by computing their expected separation at
their closest point of approach (CPA), even when the involved aircraft don’t have a
common shared point in their expected trajectories. When their trajectories share a
common 3D point in their expected trajectories, the time of arrival’s interval at the
junction can also be used. Any A/C following a predefined flight plan, as defined in their
RBT, usually have relevant along track unpredicted deviations, leading to strong
variations on the location and separation at their CPA, whereas the common 3D point
remains unchanged. Hence, a definition of “time interval” between arriving A/C TOAs at
a junction has been chosen to identify and to mitigate potential conflicts in this paper.
The minimum required “safe” time separation (𝜏) between A/C at the crossing
point or junction, defined at strategic level, depends on the degree of the adherence of the
actual to the planned trajectory. Eurocontrol29
defines the planned trajectory as the most
likely behaviour of a flight through an area of interest over medium and long term
planning horizons. As suggested in the definition, this trajectory may suffer from various
sources of uncertainties as it is calculated on the basis of assumptions of expected A/C
behaviour and, meteorological and other conditions. These uncertainties involve vertical,
lateral (cross-track) and longitudinal (along-track) deviations. These deviations are
assumed here as statistically independent or uncoupled.
Additionally, the uncertainties (due to initial time or scheduling) also affect the
A/C TOA at the crossing point. The following sections describe and quantify all different
sources of these uncertainties.
4.5.1 Effect of cross track errors
In the PBN concept, the lateral or cross-track deviation from a nominal trajectory
is due to the Total System Error (TSE). It causes deviations on the actual junction
coordinates. To establish the contribution of these errors on the time separation
deviations, Figure 6 shows an example of two A/C; A/C 𝑗 and A/C𝑖, each assumed flying
in a straight trajectory.
The solid lines represent the trajectory centrelines whilst the doted lines represent
the lateral deviations. The ideal trajectories for the A/C 𝑖 and 𝑗 can be expressed
respectively by the following equations:
𝑦 = 𝑎𝑖 ∙ 𝑥 + 𝑏𝑖 𝑦 = 𝑎𝑗 ∙ 𝑥 + 𝑏𝑗
Since, at the intersection point, the two equations have the same coordinates, by
equating the two equations and solving for x and y, then:
𝑥 =𝑏𝑗−𝑏𝑖
𝑎𝑖−𝑎𝑗
𝑦 =𝑎𝑖𝑏𝑗−𝑎𝑗𝑏𝑖
𝑎𝑖−𝑎𝑗
Figure 3.Geometrical uncertainties at the crossing point
It can be noted that the lateral deviations are due to changes in the value of
coefficients 𝑏𝑖 and 𝑏𝑗 of trajectories equations (𝛿𝑏𝑖,𝛿𝑏𝑗), moving the junction coordinates
(𝛿𝑥,𝛿𝑦) within an uncertainty area. The deviations due to these variations can be
expressed in the following matrix form:
[𝛿𝑥𝛿𝑦
] = [
−1
𝑎𝑖−𝑎𝑗
1
𝑎𝑖−𝑎𝑗
−𝑎𝑗
𝑎𝑖−𝑎𝑗
𝑎𝑖
𝑎𝑖−𝑎𝑗
] ∙ [𝛿𝑏𝑖
𝛿𝑏𝑗] (11)
Based on this result, the covariance matrix for the junction coordinates error
(𝛿𝑥,𝛿𝑦) can be derived from the (𝛿𝑏𝑖,𝛿𝑏𝑗) deviations. Multiplying matrix [𝛿𝑥𝛿𝑦
] in
Equation (11) by its transpose matrix [𝛿𝑥 𝛿𝑦] results in:
[𝛿𝑥𝛿𝑦
] ∙ [𝛿𝑥 𝛿𝑦] =
[
−1
𝑎𝑖 − 𝑎𝑗
1
𝑎𝑖 − 𝑎𝑗
−𝑎𝑗
𝑎𝑖 − 𝑎𝑗
𝑎𝑖
𝑎𝑖 − 𝑎𝑗]
∙ [𝛿𝑏𝑖
𝛿𝑏𝑗] ∙ [𝛿𝑏𝑖 𝛿𝑏𝑗] ∙
[
−1
𝑎𝑖 − 𝑎𝑗
−𝑎𝑗
𝑎𝑖 − 𝑎𝑗
1
𝑎𝑖 − 𝑎𝑗
𝑎𝑖
𝑎𝑖 − 𝑎𝑗]
Which gives:
[𝛿𝑥2 𝛿𝑥𝛿𝑦
𝛿𝑥𝛿𝑦 𝛿𝑦2 ]=1
(𝑎𝑖−𝑎𝑗)2 [
−1 1−𝑎𝑗 𝑎𝑖
] ∙ [𝛿𝑏𝑖
2 𝛿𝑏𝑖𝛿𝑏𝑗
𝛿𝑏𝑖𝛿𝑏𝑗 𝛿𝑏𝑗2 ] ∙ [
−1 −𝑎𝑗
1 𝑎𝑖] (12)
The covariance matrix is defined as the expectation for the above equation and becomes:
[𝜎𝑥
2 𝜎𝑥𝑦
𝜎𝑥𝑦 𝜎𝑦2] =
1
(𝑎𝑖−𝑎𝑗)2[−1 1−𝑎𝑗 𝑎𝑖
] ∙ [𝜎𝑏𝑖
2 0
0 𝜎𝑏𝑗
2] ∙ [−1 −𝑎𝑗
1 𝑎𝑖] (13)
Taking x-axis as coincident with the trajectory of the A/C 𝑖 with 𝑎𝑖 = 0 gives:
𝜎𝑏𝑖= 𝜎𝐿𝐷𝑖
(14)
𝜎𝑏𝑗=
𝜎𝐿𝐷𝑗
cos (𝛼𝑗) (15)
where 𝛼𝑗 is the angle between trajectories.
The variance for the position uncertainty is derived from the trace of the
covariance matrix (Equation (13)), such that the standard deviation for the position of the
junction results into a function of the standard lateral deviation of trajectories (𝜎𝐿𝐷𝑖,𝜎𝐿𝐷𝑗
)
assumed equals (𝜎𝐿𝐷) and the angle between them (𝛼𝑗) given by:
𝜎𝐶𝑃 =𝜎𝐿𝐷
sin (𝛼𝑗)√2 (16)
From Figure 6, the cross track uncertainty on the junction position obtained in
terms of distance (2𝜎𝐶𝑃), can be converted into a time standard deviation caused by this
position deviation such that:
𝜎𝑇1𝑖 =𝜎𝐶𝑃
𝑉𝑖 (17)
𝜎𝑇1𝑗 =𝜎𝐶𝑃
𝑉𝑗 (18)
4.5.2 Vertical uncertainties
A Particular interest is required for those crossing points, where involved A/C are
crossing their altitudes (or flight levels) around their 2D junction location. As it is well-
known, vertical A/C evolution contains a broad range of sources of uncertainty
variability, caused by atmosphere states, A/C performance/weight and efficiency criteria
applied by the airline operator.
To cope with this situation, whilst maintaining a conservative approach, it is
postulated in this paper that the involved A/C in this situation will be considered as if
they were flying at the same altitude at the junction.
4.5.3 Initial time (Scheduling) uncertainty
Figure 7 shows the scheduling uncertainties, expressed as initial time errors.
Considering their standard deviations, say 𝜎𝑇2𝑖 and 𝜎𝑇2𝑗, these values are translated
directly with the same values to the crossing point.
Figure 4.Translation of lateral deviations to time deviations at CPA
4.5.4 Along-track time uncertainties at the crossing point
The along-track uncertainty is due to the actual ground speed being different from
the expected speed (𝑉0). This along-track error drifts linearly with time. In order to
derive the impact of a long-track time errors, it is assumed in this paper that the distance
(𝑑) to the crossing point is known. Then, the expected flight time (𝑡0) to this point is
given by:
𝑡0 =𝑑
𝑉0
Any deviation (𝛿𝑉) in the ground speed (𝑉0) will involve a change in the arrival
time, given by:
𝛿𝑡 = −𝑑∙𝛿𝑉
𝑉2 ≈ −𝑑∙𝛿𝑉
𝑉02 (19)
From this result, the standard deviation σT3 for the TOA (𝛿𝑡) to the junction, due
to errors into the ground speed, having standard deviation of σv, can be directly obtained:
𝜎3𝑇 =𝑑
𝑉02 ∙ 𝜎𝑉 (20)
In summary, A/C 𝑖 and 𝑗 trajectory lateral deviations, initial time and along-track
time deviations are transferred as time uncertainties to the junction with values for their
standard deviations shown in Table 3
Table 3. Summary of sources of uncertainties expressed as standard deviations
Uncertainty at the junction A/C I, j TOA standard deviation
Due to A/C lateral deviation (𝜎𝑇1𝑖,𝑗) 𝜎𝑇1𝑖,𝑗 = √2 ∙𝜎𝐿𝐷
V𝑖,𝑗 ∙ sin (𝛼𝑖𝑗)
Due to Initial time deviation (𝜎𝑇2𝑖,𝑗) 𝜎𝑇2𝑖,𝑗
Due to a long-track time deviation (𝜎𝑇3𝑖,𝑗) 𝜎𝑇3𝑖,𝑗 =𝑑𝑖,𝑗
𝑉𝑖,𝑗2 ∙ 𝜎𝑉𝑖,𝑗
4.5.5 Criteria for pairwise time separation at the junction
Based on previous discussion, considering those three sources of errors being
statistically independent and assuming Gaussian distribution applied to all of them, the
required time interval (𝜏) can be derived from a predefined expected probability of ATC
intervention (Pc) at the junction .
The total standard deviation on the TOA can be computed for A/C 𝑖 and 𝑗 as:
𝜎𝑇𝑖,𝑗 = √𝜎𝑇1𝑖,𝑗2 + 𝜎𝑇2𝑖,𝑗
2 + 𝜎𝑇3𝑖,𝑗2 (21)
The expected probability of ATC intervention (Pc) can be computed by
convolving the two associated probability density functions (pdf) for A/C (𝑖, 𝑗) for the
time of arrival (𝑡) to the junction for each of them given by:
𝑓𝑖(𝑡) =1
√2𝜋𝜎𝑇𝑖𝑒
−(𝑡−𝑡𝑖)
2
2𝜎𝑇𝑖2
, and 𝑓𝑗(𝑡) =1
√2𝜋𝜎𝑇𝑗𝑒
−(𝑡−𝑡𝑗)2
2𝜎𝑇𝑗2
This convolution for both pdfs gives the probability of A/C 𝑖, 𝑗 arriving at the
same time:
𝑃𝐶 =1
√2𝜋(𝜎𝑇𝑖2+𝜎𝑇𝑗
2)𝑒
−(𝑡𝑖−𝑡𝑗)2
2(𝜎𝑇𝑖2+𝜎𝑇𝑗
2)= 1
√2𝜋(𝜎𝑇𝑖2+𝜎𝑇𝑗
2)𝑒
−𝜏𝑝
2
2(𝜎𝑇𝑖2+𝜎𝑇𝑗
2) (22)
The required time interval (𝜏𝑝) between their expected TOA for a given Pc then
results in:
𝜏𝑝 = √−2(𝜎𝑇𝑖2 + 𝜎𝑇𝑗
2)𝐿𝑛[𝑃𝐶 ∙ √2𝜋(𝜎𝑇𝑖2 + 𝜎𝑇𝑗
2)] (23)
If the expected (nominal) time interval (𝜏0) is equal or greater than the above
computed value (𝜏𝑝), there will be no additional time interval required. Otherwise, the
demanded time increment shall be:
𝜏 = 𝜏𝑝 − 𝜏0 (24)
4.6 Junction inbound capacity and TOA uncertainties
Figure 8 Shows the required TOA interval (𝜏𝑝) for different global standard
deviations (√(𝜎𝑇𝑖2 + 𝜎𝑇𝑗
2)) in the TOAs for A/C 𝑖 and 𝑗 at the junction with different
chosen expected probability of ATC intervention (Pc).
By assuming a constant TOA interval (𝜏𝑝) between consecutive A/C, the
maximum inbound flow at the junction (the frequency of traffic) can be directly derived
as:𝑄𝐼𝑚 = 1/𝜏𝑝. It can be derived from Figure 8 that in order to maintain expected
probability of ATC intervention below 10-3
and a junction capacity close to 6 A/C per
hour, a global TOA uncertainty of 3 minutes or less will be required.
Figure 8.Required TOA interval for different collision probabilities and TOA standard deviation for the
involved A/C
The results above quantitatively indicate the strong sensibility of the required
TOA interval to the A/C TOA uncertainties. This also concludes that the proposed
planned strategic changes for these TOAs allow a maximum rate of up to 6 A/C of
inbound traffic flow per hour only when the actual A/C TOA uncertainty is reduced
below 3 minutes of the standard deviation.
As discussed below, however, this TOA standard deviation value can only be
achieved under specific operational and A/C capabilities conditions.
Uncertainties due to lateral deviations, given in terms of standard deviation
(𝜎𝑇1𝑖,𝑗), as expressed in Table 3, depends on the navigation performance accuracy (given
by 𝜎𝐿𝐷 or RNP), A/C ground speed (V𝑖,𝑗), and angle between A/C trajectories/tracks
(𝛼𝑖𝑗). This dependency is shown in Figure 9. As can be observed, for crossing angles
𝛼𝑖𝑗 ≤ 200 and speeds V𝑖,𝑗 ≤ 200𝑘𝑡 the resulting relative TOA standard deviation is:
𝜎𝑇1𝑖,𝑗/𝜎𝐿𝐷 ≥ 1𝑚𝑖𝑛/𝑁𝑀.
Assuming for a typical commercial A/C with a speed greater than 200 knots and
the crossing angle between tracks greater than 200 then, under these conditions, the
relative TOA standard deviation is: 𝜎𝑇1𝑖,𝑗/𝜎𝐿𝐷 < 1𝑚𝑖𝑛/𝑁𝑀.
Figure 9.Resulting relative TOA standard deviation vs. A/C speed and tracks crossing angle
The accuracy criteria for RNP-X involves a standard deviation of 𝜎𝐿𝐷 = 𝑋/2.
This means that when an A/C is flying under PBN with RNP-X procedures, the
associated standard deviation is (𝑋/2). For instance, RNP1 involves a standard deviation
of 𝜎𝐿𝐷 = 0.5𝑁𝑀. This then results in TOA standard deviation; 𝜎𝑇1𝑖,𝑗 <𝜎𝐿𝐷1𝑚𝑖𝑛
𝑁𝑀= 30𝑠.
Concerning the initial/scheduling time deviations ( 𝜎𝑇2𝑖,𝑗), this variable can hardly
be known a-priori, especially due to the limited runway capacities constraining
departures. According to the Eurocontrol PRR 2014,30
0.9 minutes per departure due to
local ATC departure delays at the gate and 3.5 minutes delay per departure due to
additional taxi‐out time were registered at the top 30 busy airports in Europe in 2014.
These delays can be reduced by applying new operational concepts such as A‐CDM31
at
the airport by synchronising activities of all involved players to maximise the departure
predictability. Other ongoing research projects such as the UDDP31
seeks to significantly
reduce the level of these inefficiency to a target of about 30 seconds.31
Although the
previous average values sensitivities are not known, in this paper, an initial/scheduling
time standard deviation of around 1 minute ( 𝜎𝑇2𝑖,𝑗 = ±1 minute) has been adopted.
The last source of uncertainty is the along track deviation (𝜎𝑇3𝑖,𝑗). This depends
on the A/C ground speed errors (expressed by 𝜎𝑉𝑖,𝑗), the distance flown (𝑑𝑖,𝑗) and the
A/C ground speed (𝑉𝑖,𝑗). 12 kts has been often used as a typical standard deviation for
ground speed errors,32
with this value, Figure 10 shows the impact of A/C nominal
ground speed (𝑉𝑖,𝑗) and the distance to junction (𝑑𝑖,𝑗) on the along track deviation
(𝜎𝑇3𝑖,𝑗).
Figure 10.Resulting along track standard deviation vs. A/C speed and distance to the junction for 12 knots of
ground speed uncertainty standard deviation
When the distance flown to the junction is above 500NM, the along track standard
deviation grows above two minutes for A/C’s speeds below 400 knots. These results
clearly show that along track deviations reach an unacceptable level of uncertainty for
many practical flight plans, reducing the junction’s “safe” capacity to below reasonable
operational level.
To overcome these problems, it is assumed that the A/C are equipped with an on-
board CTA functionality with the accuracy of ± 30 seconds. It has been suggested that the
use of this CTA accuracy value is more effective for dynamic Demand & Capacity
Balancing than that of TTO/TTA accuracy (±3 minutes).32
However, even relaxing the
CTA accuracy (𝜎𝑇3𝑖,𝑗 = ± 30 seconds) to 𝜎𝑇3𝑖,𝑗 = ±1 minute, an acceptable result
can be achieved.
In summary, it has been pointed out how a junction inbound flow capacity
strongly depends on TOA uncertainties (Figure 8). Only when these are below a few
minutes, the expected flow capacity level remains enough to support the proposed
application of subliminal changes on the planned A/C TOA to the junctions. Table 4
specifies the adopted values of requested TOA standard deviation, based on the
assumptions made above.
The resulting final combined TOA standard deviation ( 𝜎𝑇 = 1.5 minutes ) is
used to re-establish the required TOA interval for different collision probabilities. For
these TOA standard deviations, the minimum time interval between consecutive A/C,
derived from Equation (23), is presented in Figure 11 for different probabilities of
collision (PC).
Table 4. Summary of assumed Required TOA standard deviations
Uncertainty at the junction
A/C TOA standard deviation specification
Required condition
A/C lateral deviation 𝜎𝑇1𝑖,𝑗 = 30𝑠 Crossing angle
𝛼 ≥ 200 and RNP1
Initial time deviation 𝜎𝑇2𝑖,𝑗 = 1𝑚𝑖𝑛 -
Along-track time deviation 𝜎𝑇3𝑖,𝑗 = 1𝑚𝑖𝑛 A/C CTA equipped
Combined time deviation 𝜎𝑇 = 1.5𝑚𝑖𝑛 All of the above
As shown in Figure 11, the expected probability of ATC intervention of 10-5
requires a minimum time interval of around 9 minutes, which permits the junction’s
inbound traffic flow of up to 6A/C an hour. This value for the collision risk is then
retained, by considering it will strongly reduce the probability of ATC tactical
intervention to remove conflicts.
Figure 11.TOA interval for different probabilities of collision at the junction
4.7 Criteria for A/C time separation at the junctions along trajectories
So far, our discussion has considered a single crossing point of pairwise
separation problem. In this section, the problem with multiple junctions along the
trajectory is discussed.
Any link, which represents an A/C planned trajectory (𝑖) can have different
junctions where some TOA changes are required. Additionally, more than two A/C might
demand changes in their expected arrival times at the same junction. Figure 12 shows the
distance/time evolution for A/C (𝑖) having two consecutive junctions (𝑚,𝑚 + 1) where
changes in the expected TOA are required for A/C 𝑖 and 𝑗 at junction 𝑚, and for A/C 𝑖, 𝑘
and 𝑙 at junction 𝑚 + 1.
Figure 12.Distance/time evolution for A/C i, having conflicts at nodes 𝒎 and 𝒏
To maintain the global time performance, a set of new TOA (𝑡+𝑚𝑖) at each
junction is chosen to have the same mean time (tm0) as the one for the times before
applying speed changes (𝑡−𝑚𝑖), that is to say:
𝑡𝑚0 =∑ 𝑡−
𝑚𝑖𝑖
𝑁𝑚=
∑ 𝑡+𝑚𝑖𝑖
𝑁𝑚 (25)
The above closure condition assumes that the average traffic flowing at this
junction is low enough to allow finite time changes (𝑡+𝑚𝑖).
For any junction (𝑚), the “junction equation” derived below provides the new
TOA. In this problem definition, each A/C is assigned with a subscript with the number
of the order they arrive to the junction (e.g., 1,2, ..). The time interval (𝜏𝑝𝑞) between any
two consecutive A/C (say, 𝑝 and 𝑞) can be derived from Equations (23) and (24) with the
following set of equations applied:
𝑡+𝑚2 − 𝑡+
𝑚1 = 𝜏𝑚21
𝑡+𝑚3 − 𝑡+
𝑚2 = 𝜏𝑚32
𝑡+𝑚𝑛 − 𝑡+
𝑚(𝑛−1) = 𝜏𝑚(𝑛−1)𝑛
𝑡+𝑚1 + 𝑡+
𝑚2+. . 𝑡+𝑚𝑛 = 𝑛 ∙ 𝑡𝑚0 (Closure condition)
These equations allow us to compute the new (𝑛) TOA at junction (𝑚). The above
system has to be solved over “all (active) junctions in the network” that are demanding a
new TOA.
Once all new times (𝑡+𝑚𝑖) have been computed for each node (𝑚) and for each
involved A/C (𝑖) (see Figure 13), the required speed changes can be easily established.
Let’s define the following variables and constants:
𝑉𝑚𝑚+1; Speed between junctions 𝑚 and 𝑚 + 1,
𝑑𝑚𝑚+1; Distance between junctions 𝑚 and 𝑚 + 1 (constant),
∆𝑡𝑚𝑚+1; Time of travel between junctions 𝑚 and 𝑚 + 1,
These variables are inter-related by the following equation:
𝑉𝑚𝑚+1 =
𝑑𝑚𝑚+1
∆𝑡𝑚𝑚+1 (26)
Variations on the speed between junctions (𝑉𝑚𝑚+1) are related to the variations on
the travel time (∆𝑡𝑚𝑚+1) by:
𝛿𝑉𝑚𝑚+1 = −
𝑑𝑚𝑚+1
(∆𝑡𝑚𝑚+1)2
𝛿(∆𝑡𝑚𝑚+1) (27)
Where the variations on the travel time (𝛿(∆𝑡𝑚𝑛 )) are known and given by:
𝛿(∆𝑡𝑚𝑚+1) = (𝑡+
𝑚+1 − 𝑡−𝑚+1) − (𝑡+
𝑚 − 𝑡−𝑚) (29)
Figure 13.Nominal and perturbed distance/time evolution for A/C i
In other words, all the speed changes (𝛿𝑉𝑚𝑚+1) can be directly computed by
applying the following equation to all trajectories (𝑖) where some amount of speed
change is required:
𝛿𝑉𝑖𝑚𝑚+1 = −
𝑑𝑖𝑚𝑚+1
(∆𝑡𝑖𝑚𝑚+1)2
[(𝑡+𝑖(𝑚+1) − 𝑡−
𝑖(𝑚+1)) − (𝑡+𝑖𝑚 − 𝑡−
𝑖𝑚)] (30)
4.8 Optimising the speed changes with multiple junctions
When the above closure condition (Equation (25)) was imposed, the amount of
required speed changes demanded of the different involved A/C was ignored; this may
lead to some unrealistic solutions. To overcome this risk, the closure condition is
removed and replaced by a linear optimisation problem where the cost function
represents the global speed changes for the involved A/C population. From Equation 30,
the speed change in segment (𝑚 to 𝑚 + 1) for flight (𝑖) can be expressed as a function of
the required time changes at both nodes:
𝜏𝑖(𝑚+1) = (𝑡+𝑖(𝑚+1) − 𝑡−
𝑖(𝑚+1))
𝜏𝑖𝑚 = (𝑡+𝑖𝑚 − 𝑡−
𝑖𝑚) (31)
The longer the distance along which this speed change is applied, the worst flight
efficiency will be for the involved A/C. Then the “cost” for this particular perturbation
can be modelled by the product of: 𝛿𝑉𝑖𝑚𝑚+1 ∙ 𝑑𝑖𝑚
𝑚+1, which leads to a global cost function
obtained as sum of these elementary contributions given by:
𝐽 = ∑∑𝛿𝑉𝑖𝑚𝑚+1 ∙ 𝑑𝑖𝑚
𝑚+1
𝑚𝑖
= −∑∑[𝑑𝑖𝑚
𝑚+1
∆𝑡𝑖𝑚𝑚+1]
2
[𝜏𝑖(𝑚+1) − 𝜏𝑖𝑚] (32)
𝑚𝑖
= −∑∑[𝑉𝑖𝑚𝑚+1]2[𝜏𝑖(𝑚+1) − 𝜏𝑖𝑚]
𝑚𝑖
This cost function is subject to the following conditions:
At each junction, the 𝑛 number of arriving A/C shall follow:
𝑡+𝑚2 − 𝑡+
𝑚1 ≤ 𝜏𝑚21
𝑡+𝑚3 − 𝑡+
𝑚2 ≤ 𝜏𝑚32
𝑡+𝑚𝑛 − 𝑡+
𝑚(𝑛−1) ≤ 𝜏𝑚(𝑛−1)𝑛 (33)
Furthermore, all relative speed changes (𝛿𝑉𝑖𝑚𝑚+1) shall be below a given
maximum (X %). Additionally, the total flight time shall be maintained, then for the last
junction or node (𝑘) of each flight (𝑖), the final arrival time shall be maintained.
Therefore, for all affected A/C (𝑖), the following conditions applies
𝛿𝑉𝑖𝑚𝑚+1 ≤ 0.0𝑋 ∙ 𝑉𝑖𝑚
𝑚+1
𝑡+𝑘𝑖 = 𝑡−
𝑘𝑖
(34)
The above cost function with the associated constraints should provide an
optimised set of proposed speed changes, removing all expected conflicts where the
required TOA’s changes are sufficiently small to maintain the total flight time by
applying slight speed changes for all involved A/C.
4.9 Examples for merging conflict removal at single/multiple Junctions by applying TOA subliminal changes
Different scenarios based on the number of A/C involved in the conflict at the
junction, flight distance to the junction, initial time separation interval between any two
consecutive A/C at the junction (𝜏0) and the number of junctions (single junction or
multiple junctions) are considered in the examples below.
Equation (24) considers 𝜏0 as the time separation interval between any two
consecutive A/C at the junction before the subliminal speed changes are applied. This
variable represents the interdependency of time stamps on A/C trajectories or the A/C
closeness at the junction inherently in the flight plans.
For each scenario, the expected probability of ATC intervention for any two
consecutive A/C is set to10−5, hence its resulting required minimum TOA interval ( 𝜏𝑝)
will be 9.5 minutes. The optimization model is applied for each scenario to obtain the
required subliminal TOA change for each A/C providing this expected probability of
ATC intervention. The results are then verified using simulations in the ATC2K
simulator software.
Example1: Single Merging Junction. Consider six merging A/C at a single
junction. It should be recalled that this value (6 A/C/hour) has been determined as close
to the maximum arriving rate to a junction to achieve an expected probability of ATC
intervention of 10−5 , other words, minimum time interval (𝜏𝑝) of 9.5 minutes. In this
simulation example, the optimal speed changes are calculated for each A/C before and
after the junction in order to maintain the required TOA interval.
All A/C in this example fly with the same speed of V=450 knots for a flight
distance of 900NM to the junction. The time separation interval between any two
consecutive A/C (𝜏0) value is allowed to randomly vary within different intervals where
each interval represents a different initial time stamp interdependency of A/C trajectories
at the junction before TOA changes are applied. For each 𝜏0 interval between A/C, a
different scenario is generated.
Figure 14, shows box and whisker plots of the speed changes (in percentage of
A/C speed) obtained by the optimisation model to remove merging conflict scenarios
for 𝜏0 intervals changing randomly within the ranges [0-9], [0-8], [0-7], [0-6], [0-5]
minutes and [0-4] minutes respectively, which represent scenarios above the throughput
of the junction.
Figure 14.Statistics for the required speed changes at Junction for different randomized time separation
intervals between consecutive A/C, for a set of 6A/C in bound a junction
As shown in Figure 14, for random 𝜏0 variations within intervals of [0 9] and [0
8] minutes, the obtained speed changes are all below 6%, for [0 7] minutes interval, still
the 50th
percentile is under this threshold.
The conflict situation in Figure 14 before and after TOA changes are applied is
analysed through a series of simulations using ATC2K simulator, the results are shown in
Figure 15. The left hand side shows the situation before, while its right hand side shows
the situation after TOA changes are applied.
Figure 15.Conflict removal for 6A/C merging to the Junction
Example 2: Multiple Junctions. As previously discussed in this paper, A/C may
be in conflict at multiple junctions along its flight. For example Figure 16, shows 8
flights where the A/C denoted as A/C4 is involved in a conflict with other five A/C at
junction ‘m’ and also with other two A/C at junction ‘n’.
Figure 16.Conflict on multiple junctions involving 8A/C
Using the same traffic sample as used in the previous example but adding two
A/C at the second junction, the optimisation model provides the required speed changes
that resolve all the conflicts at both junctions providing the following results (see Table
5).
Table 5. Optimal speed changes for Multiple Junction Conflict Involving 8 A/C
𝜏0 variation intervals(Minutes)
8-9
7-9
6-9
5-9
4-9
3-9
A/C First Joint
Optimal Speed Changes (% of A/C’s
Speed)
A/C1 3.2731 4.5429 5.1760 8.9412 10.8340 13.1296
-2.5738 -3.5542 -5.4552 -8.2804 -10.8359 -13.3052
A/C2 2.7878 5.3553 5.9990 7.6717 7.6493 9.6664
-2.8995 -5.4498 -5.9438 -7.1567 -7.0368 -8.3386
A/C3 0.0574 2.3431 3.5855 2.3202 4.4268 1.9553
-0.0694 -2.7684 -4.0640 -2.4374 -3.6789 -2.9741
A/C4 -1.4753 -1.2932 -0.2656 -2.5493 -1.9951 -3.9895
2.0553 1.7668 0.3484 4.2258 2.4911 4.5226
A/C5 -2.1186 -1.9689 -4.9123 -4.4023 -5.2166 -7.0531
3.3740 3.0780 4.0083 7.4302 7.5429 9.6725
A/C6 -2.1179 -4.7596 -6.1979 -7.6655 -9.7612 -10.4172
3.8212 5.2369 5.5880 8.9412 10.4107 13.5992
Second Joint
A/C7 0.2088 0.6289 0.6289 0.6289 2.3454 0.2088
-0.4149 -1.2346 -1.2346 -1.2346 -4.3825 -0.4149
A/C8 -0.5419 -0.9138 -1.6578 -0.9138 -0.9430 -0.5419
1.2658 2.1277 3.0961 2.1277 2.1277 1.2658
A/C4 -1.8405 -1.4374 -1.4374 -1.7065 -0.7581 -3.9360
Figure 17 shows the statistics for the speed changes. It can be observed that speed
changes are still below the 6% for randomised TOAs intervals between [0 9] and [0 8]
minutes in this multiple junction situation.
Figure 17.Statistics for the required speed changes to resolve conflict at multiple conflicted junctions for
different randomized time separation intervals between consecutive A/C, involving a bunch of 6 A/C arriving at
the first junction and 3A/C arriving A/C at the second junction
The performance of the proposed method provides the expected results and,
although the optimization tool has scalability problems when applied to high traffic
environments, as a LP is applied, the foreseen execution times are not expected to be a
limiting factor.
4.10 Complexity at the junctions
Strategic a priori removal of expected conflicts among planned A/C trajectories,
identified and mitigated in a centralised manner, does not ensure “free of conflict” to the
actual flown trajectories. This can be due to the appearance of additional unexpected
events (not embedded into the previously considered stochastic characterisation) or
whether because the current situation is beyond the chosen limiting boundaries (outliers).
Under any of the above situations the ATC system has to be able to maintain the
required safety levels by applying reactive actions. This capability strongly depends on
the available resources devoted to cope with them at that time and on the complexity of
the new scenario. The complexity of the airspace is changing in time and location, and
therefore, it has to be continuously monitored and mitigated, helping to employ the ATM
capabilities to properly react in the most safe and efficient manner.
Modelling the air traffic complexity is beyond the scope of this paper. The
concept used in this work is taken from previous studies.33,34
From these two studies, two
main concepts are retained here for further elaboration; air “traffic density” and “intrinsic
complexity”.
The pursued purpose is to associate a complexity indicator to any (active)
junction, generated by the involved and surrounding A/C under nominal and even
reference conditions. This assessment completes the information required to support a
planning process within traffic flow management. In other words, it will reveal whether
the involved junction should be declared a “hot spot” (that is; showing high complexity)
and subsequently it will then be considered as a situation requiring special ATM attention
and resources.
The definition of these two parameters uses the junction’s coordinates (3D), and
additionally the “time of interest”. The latter (referred as 𝑡𝑚0 for junction 𝑚) is
established as mean time among all modified (by small speed changes) expected TOA to
the junction (𝑡+𝑚𝑖) for the A/C (𝑖) among all the involved arriving A/C (𝑁𝑚) (Here it has
been assumed that a finite and small number of A/C are arriving to the junction with a
time interval below the minimum required):
𝑡𝑚0 =∑ 𝑡+
𝑚𝑖𝑖
𝑁𝑚 (35)
Once defined both, coordinates and time of interest (𝑡𝑚0), the traffic density (Dm)
and the intrinsic complexity (Im) are computed for time 𝑡𝑚0. The traffic density is defined
as superficial relative density, this changes the approach proposed by Delahaye and
Puechmorel35
to:
𝐷𝑚 = 𝑅2 ∑1
max (𝑑𝑚𝑛, 𝑅/2)2
𝑁
𝑛=1
(36)
Where 𝑑𝑚𝑛 is the distance between the A/C (𝑛) position and the junction (𝑚) and 𝑅 is
the characteristic distance used for the definition of the surrounding area. To avoid
singularity in Equation (36) when 𝑑𝑚𝑛 → 0, then if 𝑑𝑚𝑛 ≤ 𝑅/2, the value 𝑅/2 is
considered instead. The sum shall consider all A/C (𝑁) within a given ratio (say, 3𝑅)
giving representative changes on the relative density value.
The Intrinsic complexity is adopted from the work presented by Delahaye and
Puechmorel, 36
where the air traffic surrounding the junction at that reference time (𝑡𝑚0)
is modelled as a local linear dynamical. The idea of this approach is to model the set of 𝑁
A/C trajectories in this vicinity by a flow evolving like a dynamical system defined by
the following equation:
�̇� = 𝑓(𝑋)
Where �̇� and 𝑿 are respectively the speed and position of the A/C. In the surrounding
area of the junction m, located at 𝑿𝒎 and assuming existence of the first spatial
derivatives, the equation above can be rewritten as:
𝑑�̇� = 𝐽𝑓 𝑑𝑋 (37)
Where 𝑱𝒇 represent the Jacobian matrix for the function 𝒇(𝑿) at point 𝑿𝒎. Then,
assuming smooth behaviour of 𝒇(𝑿) it can be approximated by:
�̇� − 𝑿�̇� = 𝑱𝒇(𝑿 − 𝑿𝒎) + 𝑯𝑶𝑻 ≈ 𝑱𝒇(𝑿 − 𝑿𝒎) (38)
By taking 𝑿𝒎 as the origin, Equation (38) becomes:
�̇� ≈ 𝑱𝒇 ∙ 𝑿 + 𝑿�̇� (39)
Variable 𝑋 now represents the coordinates of the A/C in the proximity of the
junction and �̇� the speed of the “flow of A/C” at this point. The normal terminology for
linear dynamical systems is used by calling A=𝑱𝒇 and B=𝑿�̇�, the homogenous part of
Equation (39), (𝑨 ∙ 𝑿) exhibits the dependence of the velocity on the A/C position,
whereas the independent part (𝑩) represents the drift or velocity for the A/C flow at 𝑿𝒎.
The real part of the eigenvalues for A matrix are related to the convergence or the
divergence property of the traffic flow at the junction. When the eigenvalue has a positive
real part, the flow in the correspondence eigenvector is considered in expansion and when
it is negative the flow is considered in contraction. Furthermore, the imaginary part
provides information on the level of curl organization of the flow dynamic.
The Least Mean Square regression is applied in order to extract the best estimate
for matrix A and vector B. Based on the observations of A/C (positions and speed
vectors), the dynamical system model is adjusted. For each A/C (𝑛), it is supposed that
the position (𝑿𝒏) and its velocity (𝑽𝒏 ) are given. An error criterion between the system
model and the observations is used to define the cost function (𝐿) to be minimised as:
𝐿 = ∑‖𝑽𝒏 − (𝑨 ∙ 𝑿𝒏 + 𝑩)‖2
𝑁
𝑛=1
(40)
From this equation, the A matrix is derived.
The eigenvalues of matrix A are used to compute the intrinsic complexity. The
greater their negative real part, the higher degree of convergence to the junction for the
existing air traffic which indicates a worse scenario. In addition, their imaginary part
represents the traffic vorticity which involves a more complex situation. When the
topology of the scenario was introduced (see Figure 1), it was assumed that internal
nodes included the whole TMAs. Traffic inside this airspace is the one where a higher
degree of vorticity are expected, therefore, in order to simplify the approach only the real
part of eigenvalues of matrix A are retained for the definition of the proposed intrinsic
complexity indicator. In the same way, the eigenvectors provide the “main directions” (in
average sense) from where the incoming traffic is arriving to the junction with a
characteristic time established by its eigenvalue (−1
𝑟𝑒𝑎𝑙(𝑒𝑖𝑔(𝑘) 𝑠, in unit of time).
The degree or time of convergence, represented by the opposite of the inverse of
the real part of eigenvalues of matrix A (−1
𝑟𝑒𝑎𝑙(𝑒𝑖𝑔(𝑘) 𝑠), is then compared to the expected
reaction time for the ATC system (TATC, which is in around few minutes). The pursued
result shall provide a relevant indicator on the ratio between the traffic convergence time
and the ATC required reaction time (TATC), representing the severity of the situation.
Therefore, −𝑇𝐴𝑇𝐶∙𝑟𝑒𝑎𝑙(𝑒𝑖𝑔(𝑘)) > 1 represents strong severity situations, whilst the
smaller ratios give a more relaxed scenario. Furthermore, the proposed intrinsic
complexity indicator (𝐼𝑚) at the junction (𝑚) is defined as an exponential function of the
real parts of the three eigenvalues for matrix A (𝑘):
𝐼𝑚 = ∑ 𝑒−𝑇𝐴𝑇𝐶∙𝑟𝑒𝑎𝑙(𝑒𝑖𝑔(𝑘))
3
𝑘=1
(38)
The final chosen complexity indicator (𝐶𝑚) for junction 𝑚 is then built by
multiplying the traffic relative density (𝐷𝑚) as defined in Equation (36) and the above
intrinsic complexity (𝐼𝑚), assuming that𝑁 ≥ 2:
𝐶𝑚 = [𝑅2 ∑1
max (𝑑𝑚𝑛, 𝑅/2)2
𝑁
𝑛=1
] ∙ [∑ 𝑒−𝑇𝐴𝑇𝐶∙𝑟𝑒𝑎𝑙(𝑒𝑖𝑔(𝑘))
3
𝑘=1
] (39)
It is important to establish a threshold to differentiate those (active) junctions
representing “hot spots” from the others representing “normal” scenarios, demanding
only small TOA changes. To this end, a characteristic distance 𝑅 = 5𝑁𝑀 and ATC
reaction time 𝑇𝐴𝑇𝐶 = 2 𝑚𝑖𝑛 are assumed.
Then, the proposed threshold is chosen as the equivalent to that used to trigger the
existence of a “traffic alert” event between two A/C within a typical ATC-RDPS-STCA
algorithm. Values for the STCA characteristic time and relative distance for the intruder
A/C are around: 𝑇𝑆𝑇𝐶𝐴 = 90𝑠 and 𝑑~15𝑁𝑀 leading to the following value for the
complexity (Equation 39):
𝐶𝑚 = [4 +1
9] ∙ [𝑒
𝑇𝐴𝑇𝐶∙1
𝑇𝑆𝑇𝐶𝐴] ≈ 16
From this result, it is considered that any junction having the complexity (𝐶𝑚)
higher than the above number will be classified as “hot spot”.
The exponential behaviour of Equation (39) regarding the convergence of traffic
(represented by the last factor), produces a very fast grow result when the characteristic
time (−1
𝑟𝑒𝑎𝑙(𝑒𝑖𝑔(𝑘)) drops below the ATC’s typical reaction time (𝑇𝐴𝑇𝐶∙). For instance,
taking the conditions for TCAS TA time (𝑇𝑇𝐶𝐴𝑆 ≈ 48 𝑠𝑒𝑐) and distance between A/C
(𝑑 ≈ 5 𝑁𝑀), the resulting complexity (𝐶𝑚) raises up to 60, almost four times the
established threshold. As a comparison, for a TCAS RA situation, the complexity value is
above 246.
The identification of a junction as “hot spot” is here interpreted as situations
where subliminal TOA changes are not enough to remove conflicts or, in other words,
indicate that some required speed changes are above any realistic value.
Further developments of this concept will complete the assessment of any TOA
change, providing the initial and resulting complexity, during the evolution of the traffic
through the junction. This information about the junction as hotspot should be used as
complementary information to finally decide the best measure to cope with the final
ATFCM measure.
5 Conclusions
Implementation of direct routes and free routing airspace, applied to high density
airspace, are changing the traffic flows patterns, forcing both, ATFCM and ATC, to
change. ATC sector occupancy and dynamic sectorisation are some steps in that
direction, but they are still anchored in the conventional concept “airspace based
operations” rather than in the new one “TBO”. This paper postulates a new metric for the
demand measure, based on hotspots and, as well, derives a method for establishing the
corrective actions to mitigate them at strategic level, fully aligned with the TBO concept.
Hotspots are here defined as “active” junctions, where a bunch (of at least two)
flights are expected to cross their trajectories with less than a well-defined minimum time
interval, demanding a special attention by ATC and, likely to produce reactive corrective
actions. Based in the initial RBTs, these hotspots are identified by the NM, this
identification includes the expected TOAs for the involved A/C. The minimum “safe”
time interval has been derived in this paper and, for a given set of conditions, is
established to be 9.5 minutes.
The ATFCM mitigation actions are based on establishing the new TOA to the
junctions for all A/C that remove conflicts, considering all potential space-temporal
interdependencies. The computation of these times is based on basic LP optimization,
where the total amount of distance-weighted speed changes are minimized and the initial
targeted departure and arrival times are maintained (as constraints). A maximum allowed
speed change is also imposed. These new times shall be issued by the NM to the A/C to
be included within the new RBT as requested target times to overfly (TTOs) for crossing
points, and target times to arrival (TTA) to TMAs entry points.
The initial results show a good performance in terms of the A/C speed changes
feasibility for the proposed TOA changes, and the complete removal of conflicts, and
then, of the corrective ATC action under ideal conditions. This behavior has been
obtained for different traffic samples, including a relevant junction’s overload (up to six
A/C arriving at the junction at a rate above the junction capacity) and spatial-temporal
interdependencies.
Further work, in terms of the complementary complexity metric, introduced at the
end of the paper, is still pending. Furthermore, the behavior of the proposed optimization
method, applied to a wider airspace and more realistic traffic sample, has to be also
analyzed.
References
1. Zhang W, Kamgarpour M, Sun D, et al. A hierarchical flight planning
framework for air traffic management. Proc IEEE. 2012;100(1):179–94.
2. Eurocontrol Performance Review Commission. “Performance Review Report-
An Assessment of Air Traffic Management in Europe during the Calendar
Year 2016. Draft Final Report for consultation with stakeholders.”
2017;(April).
3. Eurocontrol - STATFOR. 2013 (Dec.). Long-term forecast – flight
movements 2012 - 2035. Tech. rept.Eurocontrol, Brussels (Belgium). 1, 2.
4. Majumdar A, Ochieng W. Factors Affecting Air Traffic Controller Workload:
Multivariate Analysis Based on Simulation Modelling of Controller
Workload. Transportation Research Record: Transp Res Rec. 2002. 58-69 p.
5. SESAR Consortium. The Roadmap for Delivering High Performing Aviation
for Europe, European ATM Master Plan. 2015;
6. FAA. Concept of Operations for the Next Generation of Air Transportation
System. Tech Rep. 2010;
7. Nieto FJS. The long journey toward a higher level of automation in ATM as
safety critical, sociotechnical and multi-Agent system. Proc Inst Mech Eng
Part G J Aerosp Eng. 2015;0(0).
8. Richetta O, Odoni AR. Dynamic solution to the ground-holding problem in air
traffic control. Transp Res Part A. 1994;28(3):167–85.
9. Vranas PB, Bertsimas DJ, Odoni AR. The Multi-Airport Ground-Holding
Problem in Air Traffic Control. Oper Res. 1994;42(2):249–61.
10. Bertsimas D, Patterson SS. The air traffic flow management problem with
enroute capacities. Oper Res. 1998;406–22.
11. Bertsimas D, Lulli G, Odoni a. An Integer Optimization Approach to Large-
Scale Air Traffic Flow Management. Oper Res. 2011;59(1):211–27.
12. Kuchar JK, Yang LC. A review of conflict detection and resolution modeling
methods. Intell Transp Syst IEEE Trans. 2000;1(4):179–89.
13. Vela A, Solak S, Singhose W, et al. A mixed integer program for flight-level
assignment and speed control for conflict resolution. Proc IEEE Conf Decis
Control. 2009;5219–26.
14. SESAR Consortium, “The ATM target concept, milestone deliverable 3”,
Version 1.0, 2007;
15. Gawinowski G, Drogoul F, Guerreau R, et al. Erasmus contribution to the
2020 SESAR scenario. AIAA/IEEE Digit Avion Syst Conf - Proc. 2008;1–10.
16. Rey D, Rapine C, Fondacci R, et al. Subliminal Speed Control in Air Traffic
Management: Optimization and Simulation. Int J Transp Sci Technol.
2016;50(1):240-262.
17. Drogoul F, Averty P, Weber R. ERASMUS Strategic Deconfliction to Benefit
SESAR. Eur Air Traffic Manag Res Dev Semin. 2009;
18. International Civil Aviation Organisation. Time adherence in TBO.
2014;ICAO(October). Available from:
http://www.sesarju.eu/sites/default/files/documents/reports/WP636
19. SESAR P04.07.02 - D22 Preliminary OSED_4, edition. January 2016.
20. Delahaye D, Puechmorel S, Hansman RJ, et al. Air Traffic Complexity Map
Based on Non Linear Dynamical Systems. Air Traffic Control Q.
2004;12(4):367–88.
21. Gurtner G, Bongiorno C, Ducci M, et al. An empirically grounded agent based
simulator for the air traffic management in the SESAR scenario. J Air Transp
Manag. 2017;59:26–43.
22. Gatsinzi D, Nieto FJ , Madani I. “Hot Spot Identification and Mitigation at
Strategic Level by Subliminal Changes in Aircraft Time of Arrival at
Junction.” proceedings of SESAR Innovation ,November 2016, Delft , The
Netherlands.
23. iFly project.Complexity metrics applicable to autonomous A/C. Deliverable
D3.1. (2009).
24. Sáez Nieto FJ, Arnaldo Valdés R, García González EJ, McAuley G, Izquierdo
MI. Development of a three-dimensional collision risk model tool to assess
safety in high density en-route airspaces. Proc Inst Mech Eng Part G J Aerosp
Eng .2010;224(10).
25. Hilburn B. Cognitive Complexity in Air Traffic Control a Literature Review.
EEC note 2004;(04).
26. Sun D, Bayen AM. Multicommodity Eulerian-Lagrangian Large-Capacity
Cell Transmission Model for En Route Traffic. J Guid Control Dyn
.2008;31(3):616 28.
27. Cao Y, Sun D. A parallel computing framework for air traffic flow
management. Proc 2011 Am Control Conf. 2011;2771–6.
28. Day V. The ATC Interactive Radar Simulator Operations Handbook.
2002;(August).
29. Mondoloni S, Swierstra S. Commonality in disparate trajectory predictors for
air traffic management applications. In: AIAA/IEEE Digital Avionics Systems
Conference - Proceedings. 2005.
30. Commission PR. Performance Review Report. An Assessment of Air Traffic
Management in Europe during the Calendar Year 2014. 2015;(March).
31. Team CI, Manager Q, Team CI. SESAR Concept of Operations.
2007;(July):1–214.
32. Mutuel LH, Neri P, Paricaud E. Initial 4D Trajectory Management Concept
Evaluation. Eur Air Traffic Manag Res Dev Semin. 2013.
33. Delahaye D, Puechmorel S, Hansman J, et al. Air traffic complexity based on
nonlinear dynamical systems. In: ATM 2003, 5th USA/Europe Air Traffic
Management Research and Development Seminar. 2003.
34. Prandini M, Piroddi L, Puechmorel S, et al. Toward Air Traffic Complexity
Assessment in New Generation Air Traffic Management Systems. IEEE Trans
Intell Transp Syst. 2011;12(3):809–18.
35. Delahaye D, Puechmorel S. Air traffic complexity: towards intrinsic metrics.
3rd USA/Europe Air Traffic Manag R&D Semin . 2000;(June):1–11.
36. Delahaye D, Puechmorel S. Air traffic complexity based on dynamical
systems. Proc IEEE Conf Decis Control. 2010;2069–74.
Abbreviations
A/C Aircraft
A-CDM Airport Collaborative Decision Making
ANSP Air Navigation Service Providers
ATC Air Traffic Control
ATCO Air Traffic Controller
ATFCM Air Traffic Flow and Capacity Management
ATFM Air Traffic Flow Management
ATM Air Traffic Management
ATS Air Traffic Services
CARD Conflict And Risk Display
CD&R Conflict Detection and Resolutions
CNS Communication, Navigation and Surveillance
CPA Closest Point Approach
CTA Controlled Time of Arrival
CWP Control Working Position
DCB Demand Capacity Balancing
DMAN Departure Manager
E-AMAN Extended Arrival Manager
ECAC European Civil Aviation Conference
EFPL Extended Flight Plan
ERASMUS En-Route Air Traffic Soft Management Ultimate System
FMS Flight Management System
FOC Flight Operation Centres
FR Free Routing
FRA Free Routing Airspace
FTS Fast Time Simulations
GDP Ground Delay Problem
HMI Human Machine Interface
ICAO International Civil Aviation Organisation
MILP Mixed Integer Linear Programme
Navaids Navigation Aids
NM Network Manager
PBN Performance Based Navigation
PRR Performance Review Report
RBT Reference Business Trajectory
RNP Required Navigation Performance
RTA Required Time of Arrival
SESAR Single European Sky ATM Research
STAM Short Term ATFCM Measure
STCA Short Term Collision Avoidance
TBO Trajectory Based Operations
TCAS-RA Traffic Collision Avoidance System-Resolution Advisories
TCAS-TA Traffic Collision Avoidance System-Traffic Advisories
TC-SA Trajectory Control by Speed Adjustment
TMA Terminal Manoeuvring Area
TRACT Trajectory Adjustment through Constraint of Time
TSE Total System Error
UDDP user driven prioritisation process
US United States
WP Waypoint
Recommended